Once economists' consensus is out, the likeliest move came true 96.1% of the time across 1,112 decisions since 2010. Weeks earlier, from nothing but the run of decisions, 80.1%. 3 meetings out, 77.7%: and there the odds were only 6% better than guessing each move's usual share.
By Dylan Forexia · Published · updated with the data
Forecasts scored1,324ten banks, 2010 to 2026
Consensus out96.1%likeliest came true
Weeks before80.1%from the run alone
3 meetings out77.7%6% better than base rates
In brief
A week out, economists are hard to beat: the consensus alone matched 96.1% of decisions. Counting how often it has been wrong improved the forecast only a little (Brier score 0.075 against 0.077).
Weeks out, the run of decisions does most of the work: banks hold far more than they move, and they rarely turn around soon after a move.
3 meetings out, the edge is nearly gone: 6% better than guessing each move's usual share.
Every forecast was made with only the decisions before it, and every one is counted, the misses too.
How it was tested
Every decision of the Federal Reserve, the ECB, the Bank of England, the Bank of Japan, the Bank of Canada, the RBA, the RBNZ, the SNB, Norges Bank and the Riksbank from Jan 7, 2010 to Nov 3, 2026 was forecast before it was counted, using only the decisions before it, in time order. A forecast is a chance of a hike, a hold and a cut. Two rules, fixed in advance:
With economists' consensus out (in FXStreet's calendar, usually the week before): how often decisions with the same expected move went each way, the bank's own record pulled toward all ten banks' (as if all ten were 12 more of its own decisions).
Before it: how often the bank moved after the same run, which way its last move went and how many holds since. For meetings further ahead, the same rule chained over every path the meetings between could take.
The score is the Brier score, the squared distance between the forecast and what happened: 0 is perfect, and a forecast that always says each move's usual share scores the baseline. Lower is better. The same rules give the rate odds this site publishes for every meeting ahead.
A week out: the consensus is hard to beat
Economists' consensus matched the decision 96.1% of the time over 1,112 decisions with one. The 43 misses, 3.9%, are almost all of what the odds can add: they turn "a hike is expected" into "a hike, very likely", with a measured chance of something else. That took the Brier score from 0.077 to 0.075. Better, but only just: a week out, the work has been done by the people forecasting.
The latest 10 of the 43 decisions since 2010 that went another way than the consensus.
Weeks out: the run of decisions
Before the consensus is out, the run of decisions alone picked the outcome 80.1% of the time over 1,324 decisions, against a baseline Brier score of 0.362 for 0.288: 20% better. Most of that is holds.
Right after a cut, the next decision was another cut 48% of the time and a hike 0%; right after a hike, another hike 60% and a cut 0%. After six holds or more, 92% of decisions were another hold.
After
Decisions
Cut
Hold
Hike
Just cut
181
48%
52%
0%
1-2 holds since a cut
164
23%
77%
0%
3-5 holds since a cut
146
10%
86%
3%
6+ holds since a cut
546
3%
94%
4%
Just hiked
159
0%
40%
60%
1-2 holds since a hike
99
1%
74%
25%
3-5 holds since a hike
77
12%
77%
12%
6+ holds since a hike
78
14%
81%
5%
No move yet, first decision
10
30%
50%
20%
No move yet, 1-2 holds
9
11%
89%
0%
No move yet, 3-5 holds
9
22%
67%
11%
No move yet, 6+ holds
5
20%
80%
0%
All ten banks since 2008, by the run before each decision. These are the counts as they stand now; each forecast above used only the counts before it.
One meeting ahead, before the consensus3 meetings aheadPerfectly honest
3 meetings out: close to guessing
Forecasting 3 decisions ahead, the likeliest outcome still came true 77.7% of the time over 1,319 decisions, but mostly because holds are common. Against the baseline, the score was only 6% better (0.339 against 0.360). The chart shows the forecasts stayed honest, near the line, yet they had little to say: a few months out, what a bank will do depends on data that has not been published yet.
When the run alone made the odds 90% or more sure, a week or more before the consensus, they were right 500 times in 541.
Bank by bank
The Bank of Japan was the easiest to call weeks ahead (93.4%), Norges Bank the hardest (67.6%). Banks that meet more often hold more often, which flatters their figure.
How often the likeliest outcome came true. Each bank's name opens its odds for the meetings ahead.
What this does not show
These forecasts know nothing but the record: not the banks' speeches, not inflation or jobs figures, not markets. A forecaster who reads those does better weeks ahead, which is what the consensus captures a week out. Nor does a right call on the rate say what price will do: an expected move is usually in the price already, which is why surprises matter and why the reaction is measured separately.
Questions
Predicting central banks: questions
Can central bank decisions be predicted?
Mostly, a week ahead: economists' consensus matched 96.1% of 1,112 decisions since 2010. Weeks ahead, the run of decisions alone picked the outcome 80.1% of the time, mostly by calling holds. 3 meetings ahead, the forecasts barely beat base rates.
Which central bank is easiest to predict?
Weeks before the consensus, from the run of decisions alone: the Bank of Japan (93.4% of 167 decisions). Hardest: Norges Bank (67.6% of 102).
How were the forecasts made?
With fixed rules and only the decisions before each one: how often decisions with the same expected move went each way once the consensus was out, and before then how often the bank moved after the same run of decisions. The rules are on the rate odds page.
Can I quote these figures?
Yes, with a link to this page. Questions about the method go to Dylan Forexia, dylan@forexiapro.com.
How to cite
The figures on this page are free to quote, in an article, a paper, a video or an AI answer, with a link to this page:
InterestRateDecision.com, "Can central bank decisions be predicted?", updated Nov 3, 2026, https://interestratedecision.com/blog/can-central-bank-decisions-be-predicted/
The data behind them is free to download as CSV and JSON. Questions about the method: the editor.
By Forexia
Learn the read behind every projection.
Forexia teaches the same read this site measures: where the stops rest before a release, the hunt, and the close back that says it is done. Courses, live signals for members, and a community of traders who chart the same way.